Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 21 Sep 2001 11:10:30 -0400
From:      Joe Abley <jabley@automagic.org>
To:        JINMEI Tatuya / ??????????? <jinmei@isl.rdc.toshiba.co.jp>
Cc:        freebsd-net@FreeBSD.ORG
Subject:   Re: ipv6/gif/cisco syslog noise
Message-ID:  <20010921111029.H4205@buffoon.automagic.org>
In-Reply-To: <y7vadzopq9h.wl@condor.jinmei.org>
References:  <20010919153739.K85635@buffoon.automagic.org> <20010920.050441.28824742.ume@mahoroba.org> <20010919164416.Q85635@buffoon.automagic.org> <y7vadzopq9h.wl@condor.jinmei.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Sep 21, 2001 at 11:33:46PM +0900, JINMEI Tatuya / ??????????? wrote:
> >>>>> On Wed, 19 Sep 2001 16:44:18 -0400, 
> >>>>> Joe Abley <jabley@automagic.org> said:
> 
> jabley> The tunnel is configured like this:
> >> 
> jabley> buffoon# ifconfig gif0
> jabley> gif0: flags=8011<UP,POINTOPOINT,MULTICAST> mtu 1280
> jabley> inet6 fe80::2d0:b7ff:fe79:a0a7%gif0 --> :: prefixlen 64 scopeid 0x4 
> jabley> inet6 2001:438:1fff:ffff:8::32 --> 2001:438:1fff:ffff:8::31 prefixlen 126 
> >> 
> >> It should be /128.
> 
> > Problem solved. Thanks :)
> 
> Out of curiosity, why did you use /126 in the former configuration?
> We're now discussing how the kernel should treat IPv6 global addresses
> on a p2p link, and we are interested in actual users' intentions.

I spend my waking hours arm-deep in cisco and Juniper routers, and
those (and other) vendors frequently do not distinguish between
ptp and (nb)ma networks in terms of addressing.

In this case, on the cisco router which terminates the tunnel, defining
the tunnel interface with a 126-bit netmask causes a /126 prefix to be
distributed in the IGP, and this provides reachability information for
both ends of the tunnel to other routers in the network.

If the cisco interface had been numbered with a 128-bit netmask, a
/128 prefix would have been distributed which would have provided
reachability information for the cisco tunnel interface, but not the
FreeBSD gif interface. Hence traffic sourced from the FreeBSD box using
that gif interface address as a source address would not get replies
routed correctly.

My reason for using a 126-bit netmask on the FreeBSD router was
mainly due to familiarity with the cisco/juniper way of doing things.
I see that when I configure a gif interface with a 128-bit netmask
and specify a destination, I get two host routes in the kernel (one
for each end of the tunnel). This seems equally valid as far as
propagation of the routes is concerned.

The reason for allocating four addresses (a 126-bit prefix) to the
point-to-point link stems from similar practice in IPv4, I think (it
wasn't my decision; it's current policy in AS6461).



Joe

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-net" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20010921111029.H4205>